A couple weeks ago, Fran Poodry asked me what was up with words that start with a silent P, and I had to confess I didn't know. (Such words include pfennig (in the American pronunciation), pneumatic, psalm, words starting with pseud-, psi, psittaform, Psmith, words starting with psych-, ptarmigan, pterodactyl, ptomaine, and so on.) The dictionary and other reference works were (ahem) silent on the subject. Fortunately, my panel of experts was able to provide plentiful information.
The short answer is that most words that begin with a silent P in English come from Greek, where the initial /p/ sound was pronounced. For instance, the Greek word that "psalm" derives from began with the Greek letter psi, which in Classical Greek was pronounced with a /ps/ sound. But languages often restrict the sequences of sounds that are allowable in words; Latin words couldn't start with a /ps/ sound, so the Romans dropped the initial /p/ sound while retaining their transliteration of the Greek spelling. (Elliott notes that it's not clear why they took this particular approach to the illegal-sound-sequence problem. They could instead have taken other common approaches, such as dropping the second sound (the /s/) or inserting a vowel between the /p/ and /s/ sounds.) And so the words have come down to us with extra silent letters stuck on like vermiform appendixes.
(A couple of asides on pronunciation: in Homeric Greek, phi and theta and chi were not pronounced /f/, /T/, and /tS/, but rather as aspirated /p<h>/, /t<h>/, and /k<h>/, respectively. And our letter Y (like our letter U) derives from the Greek letter upsilon, which was pronounced /u/. So words that begin with "psycho-" would have been pronounced in Greek as /'psu k<h>o/ rather than /'saI ko/.)
Similarly, the somewhat obscure English words "cnidoblast" and "ctenophore" both start with a silent C. In the Classical Greek roots of those words, the initial /k/ sound was pronounced.
And speaking of initial /k/ sounds, initial K turns up silent in English as well. The modern English word "knight" was pronounced /knIxt/ in Middle English. (Where /x/ is like the German "ch" in "ich.") Both the initial /k/ and the internal /x/ apparently became silent during the transition to Modern English.
Another case of a silent written-consonant pair is the English word "chthonic," which starts with a silent "ch"; according to the dictionary, it's pronounced /'TA nIk/. (/T/ is the initial sound of "thanks.") In Greek, of course, this word started with chi and theta, and contained no silent letters.
So why do English words make some of these initial consonants silent? As with Latin, only certain kinds of sound sequences can appear in English syllables. In particular, the sounds in English syllables usually increase in sonority (which roughly means sound intensity or resonance) from the beginning ("onset") of a syllable to its nucleus (usually a vowel). (The major exception is that /s/, which is fairly sonorous, can start a syllable and be followed by less-sonorous sounds like /p/, /t/, or /k/.) For instance, the /t/ sound is not more sonorous than the /p/ sound, so /pt/ isn't allowed to start a syllable in English, so one of them has to become silent in words like "pterodactyl."
But if the two adjacent sounds in question occur in different syllables, they can both be pronounced without violating rules of sonority; thus, we pronounce both the /p/ and the /t/ in "helicopter."
A very few English words of Greek derivation lost not only the initial consonant sound but the initial written letter as well. The only such word I know of is "neume," though I suspect there are a few others. In other cases, silent letters in spellings of Greek-derived words have come and gone; for instance, "phantasm" was spelled "fantasme" in Middle English and Old French (and "fantasm" is still listed as an alternate spelling), though its Latin root started with "ph"; and "fantasy" derives from a similar Latin root that started with "ph" (and "phantasy" is listed as an alternate spelling, though not used much since the 19th century). The Greek root for both words started with phi, of course.
"Pfennig" is a special case of a silent P, not connected to Greek: the initial /pf/ is pronounced in German, as an affricate—a stop (/p/) that segues into a fricative (/f/). The dictionary pronunciation given for American English leaves out the initial /p/, but many English-speakers (especially if they know any German) pronounce the initial /p/ anyway. Initial /pf/ is not a sound that occurs elsewhere in English, though, so it's no surprise that the /p/ becomes silent. (Elliott notes that German /pf/ and /ts/ developed from /p/ and /t/ during the High German Sound Shift around 800 AD; he also supplies his favorite /pf/ word: Pfropfen, which means "stopper," as in "bottle stopper.")
As for the ptarmigan, that's a different kettle of fowl altogether. The word derives from Scottish Gaelic tarmachan. The OED says that a book published in 1684 spelled the word with an initial silent P, apparently in an attempt to model the spelling after Greek-derived medical words. Another published source in 1768 copied the silent-P spelling, which then came into general use.
I feel obliged to note, while on the subject, that ptarmigans have, according to my dictionary, "completely feathered feet," making them (as Elliott put it) rather like "avian hobbits."
Thanks to Mark-Jason Dominus, Elliott Moreton, and Arthur Evans for their aid in assembling this week's column, and to Fran Poodry for asking the original question.